Frequency-Domain Response Characteristics of Ultrasonic Signals and a Method for Identifying Buried Seeds After Sowing

During ultrasonic detection of buried seeds after sowing, variations in soil conditions can strongly affect ultrasonic signals and make seed identification difficult. In this study, soil moisture content, particle size, and compaction pressure were investigated to evaluate their effects on ultrasonic frequency-domain responses and seed-presence discrimination. The first-wave dominant frequency (Fg) and dominant-frequency amplitude ratio (k) were extracted from the received signals. Seed presence altered the spectral peak position and amplitude distribution, whereas the direction of the Fg shift varied with soil conditions, limiting its use as a standalone discrimination feature. In sieved soil, k showed the best discrimination at 10% moisture content, with an AUC of 0.951. Using the predefined threshold k = 1, the sensitivity, specificity, and overall accuracy were 94.7%, 81.3%, and 88.0%, respectively. In non-sieved field soil, the highest AUC was 0.923 at 15% moisture content, with an accuracy of 90.0% using k = 1. An additional 40 independent tests conducted at 15% moisture content and 0.50 MPa yielded a sensitivity of 95.0%, a specificity of 90.0%, and an overall accuracy of 92.5%. These results demonstrate the potential of ultrasonic frequency-domain features for identifying buried seeds under controlled soil conditions.